![]() ![]() Furthermore, during each study, the benefits, challenges, creative application solutions, and results are reported upon, analyzed, and critically discussed. In these studies, the design and development process of the application is documented in detail. These four ML applications are 1) An Anomaly Detection System in the brownfield on a Monorail Conveyor, 2) A multi-model Quality Inspection application in Laser Beam Welding with Supervised Learning, 3) A Deep Reinforcement Learning system for fully Autonomous Assembly on industrial robots, and 4) A self-service Quality Inspection Toolkit augmented with Machine Teaching and Interactive ML functions. The four studies each contain a Design Science research case on the development and design of an industrial ML application. The systematic literature review explores current research on industrial ML applications with a comprehensive quantitative and qualitative analysis. ![]() To this end, this thesis contains an extensive literature review and four independent studies on real-world applications of ML in the context of production at a large automobile OEM. This thesis ties into the current research and further contributes by exploring the development process, the challenges, and the value of libraries and frameworks in industrial ML applications. To alleviate the issues blocking further advancement in industrial ML, various research endeavors have been undertaken to better understand how ML applications work in industrial information systems, and which requirements and challenges practitioners face during development and implementation. ![]() As a result, while ML is applied to a growing number of systems and environments in industry, many aspects of its development remain unclear. However, these applications depend on several factors, such as the deployment environment, the amount of available data, and the ML approach chosen, which makes application development a non-trivial pursuit. Frequently, new scientific findings and discoveries offer new potentials for industrial ML applications. Thus, many companies have increased their investment in this technology to drive disruptive innovation in cost-effectiveness and performance. Several breakthrough applications and designs have further fueled the excitement for ML-empowered systems. Applying industrial ML to various applications in the production system has also become a popular focus of recent research efforts. By offering powerful tools for automation and intelligent decision-making, it has become a highly sought-after technology across most industries. Ranging from intelligent quality inspection to fully autonomous robots and vehicles, Machine Learning in industrial applications (industrial ML) is aligned to permanently change how production systems operate. In addition to accuracy, the performance metrics, including precision, recall, and F1-score were also improved using the proposed combinatory features.įew recent areas of research have had a more significant impact on industrial production than Machine Learning (ML). Compared to conventional statistical features, the new combination was able to enhance the classification accuracy of k-NN and SVM, respectively, from 91.3% to 99.9%, and from 94.8% to 99.7% in the case of IMS dataset, and from 94.1% to 98.5%, and from 94.7% to 98.4% in the case of Paderborn dataset. The data were categorized into large number of classes to closely indicate the actual fault type and size. Two datasets comprising the signals of run-to-failure tests were taken from Intelligent Maintenance Systems (IMS) and the Paderborn university repository. In this work, a new combination of fault descriptors, including three Hjorth's parameters, three statistical features and an entropy measure, is proposed and its effectiveness has been analyzed on classification performances of k-Nearest Neighbor (k-NN) and Support Vector Machine (SVM). The conventional statistical features used commonly in the past literature do not uniquely characterize the fault status, and yield satisfactory results only in limited cases, like those for artificial faults. The research on identification of artificially induced faults in bearing is available in abundance in the past literature, however, the diagnosis becomes more challenging when the fault evolves naturally inside the bearing, and especially when its stages need to be precisely tracked. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |